When should you use it?
- Your relationships live only in your inbox
- A contact list has been on your to-do list for months
- You work with brands, clients or collaborators who drift
Not for: Inboxes containing data you may not process with AI; Teams that already maintain a shared CRM
What do you need, and what do you get?
You need
- Mail and Notion connectors
- Your categories, such as client, sponsor or partner
- The fields you want per contact
You get
- A Notion database of contacts
- Last contact date and topic for each
- A list of people still waiting on you
How does it work, step by step?
- Decide the shapeChoose categories and fields before you start, so the database is usable.
- Connect read accessGive the agent your mail and a Notion workspace or page to write to.
- Run it in the backgroundLeave it to work through the threads while you get on with other work.
- Clean the edgesMerge duplicates, delete newsletter senders and correct mislabeled categories.
- Act on the owed listStart with the people waiting on a reply.
The prompt
Our own prompt, inspired by Tayla Burrell and written to work in Claude Cowork and Claude. Values in lime are placeholders.
Read my sent and received email from the last 12 months. Ignore newsletters, receipts and automated senders. For each real person or organization I correspond with, create a row in the Notion database [name] with: name, organization, category (client, prospect, sponsor, partner, colleague, collaborator), last contact date, what we last discussed in one sentence, relationship status, and whether I owe them a reply. Do not send or draft any emails. When done, list the ten people waiting longest on me.
Inspired by Tayla Burrell: 13 *actually* useful things you can do with Claude Cowork this weekend. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Automated senders and newsletters creep in; tell it to skip them.
- Summaries of old threads can misread status; spot-check key relationships.
Our take
Tayla Burrell's suggestion is a near-perfect background job for a desktop agent: it is almost entirely reading, the risk is low, and nobody would ever justify spending a weekend doing it by hand. The owed-a-reply column alone can repay the effort, since forgotten replies quietly cost freelancers and creators sponsorships and repeat clients. Decide the structure before it starts. Name the categories you want, such as client, prospect, sponsor, partner and collaborator, and the fields for each, so you do not spend the saved time reorganizing. Tell it to skip automated mail, newsletters, receipts and notifications, which otherwise flood the database with junk contacts. Give it read access to email only; there is no reason for this job to send anything. Expect edge cases, like one person with several addresses or an agency contact who changes, and spend a few minutes merging those afterward. Keep the Notion database private, since summaries of conversations can contain rates, disputes and personal details that should not be visible to collaborators sharing your workspace. Refresh it monthly.
Where the time goes back: A long-postponed data-entry job runs unattended and leaves you with a list of follow-ups.
FAQ
Why is this underrated?
Does it send anything?
Source and credit
Inspired by Tayla Burrell. We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- 13 *actually* useful things you can do with Claude Cowork this weekend
- Creator
- Tayla Burrell
- Published on
- taylaburrell.substack.com (Substack)
- Original date
- Jun 16, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Claude Cowork
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends or purchases.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.